SDPA backend priority (#9299)

This commit is contained in:
contentis
2025-08-13 20:53:27 +02:00
committed by GitHub
parent afa0a45206
commit 3da5a07510
4 changed files with 17 additions and 4 deletions

View File

@@ -285,7 +285,7 @@ def pytorch_attention(q, k, v):
)
try:
out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=0.0, is_causal=False)
out = ops.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=0.0, is_causal=False)
out = out.transpose(2, 3).reshape(orig_shape)
except model_management.OOM_EXCEPTION:
logging.warning("scaled_dot_product_attention OOMed: switched to slice attention")